Starbucks Plans Drive-Thru AI Tests, Raising Questions About Privacy and Accuracy

Starbucks is considering a drive-thru system that could recognize returning customers and suggest what they might order. The idea reaches beyond using AI to transcribe a menu request: it could tailor the screen before a customer speaks. That may shorten ordering, but Starbucks hasn’t explained how recognition would work, what customers could control, or whether the system would improve the experience enough to justify collecting more information.

CEO Brian Niccol described the plans on the company’s fourth-quarter 2026 earnings call. Starbucks expects to begin testing AI models for drive-thrus during fiscal 2027, which ends Sept. 28, 2027. Niccol said drive-thru customers account for roughly half of Starbucks patrons and framed the work as a way to make their orders faster, more convenient and more personal. The company hasn’t said where the tests will happen or whether they could affect menu prices.

Two jobs for AI at the speaker

One proposed use is straightforward: AI would help capture orders more accurately and efficiently while baristas focus on speaking with customers. In that setup, the software handles order entry and the employee remains available to resolve questions or correct misunderstandings. Whether that split actually speeds service depends on how often staff need to intervene and how reliably the system handles changes, customizations and noisy conditions.

The more ambitious idea is a personalized digital menu. Niccol said Starbucks is exploring ways to recognize repeat customers as they pull up and present suggestions, potentially before they begin ordering. The company hasn’t identified the technology. Facial recognition, license plate reading and detection through the mobile app are among the possibilities raised in the report, but Starbucks didn’t confirm any of them.

That missing detail matters because the recognition method determines what information the system needs and what choices customers should have. Starbucks also hasn’t explained whether customers would need to consent, how they could opt out, or what data might be retained. Those unanswered questions could shape whether personalization feels useful or intrusive.

The company already uses AI elsewhere in its operations. Green Dot Assist helps employees find drink recipes, troubleshoot equipment and coordinate staffing. Smart Queue helps sequence orders arriving through stores, drive-thrus, the app and delivery. These tools support staff and organize work behind the scenes; customer recognition would put automated systems in a more visible role during the ordering interaction.

Accuracy remains the practical test

Starbucks would enter a crowded field. McDonald’s, Taco Bell, KFC, White Castle and Wendy’s have used or tested AI for drive-thru orders. Customer feedback has been mixed, with complaints about incorrect orders, slow exchanges and bots that struggle to answer questions, according to research cited by restaurant technology site Canopy. Burger King has pulled back from using AI for drive-thru ordering after some customers left rather than place an order with a bot. It continues testing the technology while offering customers the option to order from a person.

The interface can fail in ordinary ways that matter at the point of sale. University of Washington AI researcher and associate professor Uttara M. Ananthakrishnan told CNET that voice systems may struggle with non-native and international accents because many models are trained on general US accents. Engines, weather and other background noise can make a kiosk microphone less reliable, too. A mistaken order is inconvenient; mishearing an allergy instruction could carry a more serious risk.

Previous tests show why the details matter. McDonald’s ended a drive-thru AI experiment with IBM in 2024 after the software struggled with accents and dialects. The company has since started another experiment, with an AI bot called Archy, as part of a stated $8.5 billion modernization plan. At Taco Bell, an AI system once accepted an order for 18,000 waters. These cases don’t establish how Starbucks’ system would perform, but they show that recognition accuracy and sensible handling of unusual requests need attention before a rollout.

Ananthakrishnan questioned whether the benefits are clear and said customer-facing AI can stumble on edge cases. In her view, customers may be more willing to use it if they know mistakes will be fixed quickly. That puts a human handoff at the center of the product design: customers need a simple way to ask for help, and staff need enough time and authority to correct errors.

Personalization needs clear boundaries

Some customers have also raised concerns about privacy and the possibility that a personalized menu could show different offers to different people. Starbucks says it doesn’t use surveillance pricing. A 2025 report in The Washington Post, however, said frequent users of the company’s loyalty app might receive fewer promotional deals. That report doesn’t establish why those customers saw fewer offers, but it adds to the uncertainty around how customer data could shape a future menu.

Price transparency will be one measure to watch. Starbucks hasn’t said whether suggestions would include discounts, whether prices could vary by customer, or how recommendations would be generated. Without those answers, “personalized” could mean a useful shortcut based on familiar orders—or a menu that gives customers little insight into why particular items appear.

Workers will judge the system, too

Starbucks says its technology is meant to strengthen human connection rather than replace it. That claim will be tested against both the customer experience and the working conditions of employees who must keep the system running. Since late 2021, employees have formed unions at nearly 700 US stores and called for higher pay and better staffing. Workers have also raised concerns about limited hours and short notice of shift changes.

Since Niccol became CEO in September 2024, Starbucks has cut more than 2,000 employees under its Back to Starbucks strategy, which has also simplified menus, reduced wait times and closed some stores. In August, Starbucks Workers United, which represents 12,000 baristas, called for a national consumer boycott of company-owned stores and criticized the company’s technology spending. Organizer Jason Woods told Current Affairs in 2025 that the company was exploring AI ordering while still struggling to cover shifts when employees called out.

That tension gives the drive-thru test a second measure of success. If AI takes repetitive order entry off employees’ hands and leaves them more time to help customers, it could support the service Starbucks says it wants. If it adds corrections, technical work or new staffing pressure, the promised convenience may not reach either customers or baristas. The fiscal 2027 tests should reveal whether Starbucks can make recognition accurate, give customers meaningful control and keep a person ready when the software gets something wrong.